November 2026
Stunt Double Studio, free on your own machine, plus a browser for each checklist
Studio is new and free: an actor runs your checklists in Chrome on your own machine, from your coding agent or your browser, and shows you which checks fail with screenshots. Checklists can also now run at the device size and language you choose.
- Stunt Double StudioNew
Studio runs your checklists locally, for free. An actor goes through your product in a real Chrome on localhost, staging or production, and every pass cites what it saw on the page, with a screenshot. There is no account and no Stunt Double bill: it runs on the plan or model key you already use.
- Run Studio from your agent or your browserNew
Install Studio in your coding agent or any MCP client and ask it to run a checklist, or open Studio in your browser with no agent at all. Every run is kept as plain files in your project, so you can commit, compare and share them, and move a checklist to the cloud when you want scheduled runs, an independent judge or team sharing.
- Choose the browser a checklist runs inNew
A checklist can now run as desktop, tablet or mobile, or at a custom size with touch, and in a language you pick, such as en-AU. Set it under Configuration on the checklist, and every run opens that browser.
- Map large sites soonerImproved
A topology crawl now draws the map of your site first and adds page captures after, so you can explore a large site within minutes. The map shows its captures arriving, and paid plans can map far larger sites.
- Every automation in its projectImproved
Automations made on the old workspace page now sit in the project they run against. Any we could not place show under Not in a project on each project’s Automations page, with a button to move them there.
- Runs that finishImproved
A checklist run that loses its browser part way through now reconnects and carries on, keeping the checks it already made, and fewer runs stop early on an internal error.
- Smaller fixesImproved
A checklist scored in points no longer shows a percentage on its score card. Index scorecard and compare pages load faster, and app pages that used to crash on a page update now keep working. Slow replies in chat now end with a clear message instead of hanging, a topology crawl no longer fails when one of its browser tabs is slow to open, and the public docs are clearer on how the MCP server signs you in and which plans include devices.